Major websites blocking AI agents from performing tasks, limiting utility of autonomous assistants
OpenAI, Meta, Google, and other AI companies are rolling out autonomous agents to conduct online tasks on users' behalf, but popular websites are increasingly blocking them, raising questions about their practical viability.
What happened
In recent weeks, users of AI agents from Meta (Muse), xAI (Grok), OpenAI (Dots), and Instinct have reported blocks preventing these agents from performing tasks on major websites. Yelp restricts AI agents to those paying for its data licensing program; Delta Air Lines prevents them from booking tickets; JPMorgan Chase and Charles Schwab block agent access due to security concerns. Amazon has blocked multiple agents including Meta's Muse and reports indicate Cars.com and Amtrak have also quietly restricted access. This follows Amazon's 2024 legal action against Perplexity's shopping agent, though an appeals court overturned an initial injunction in August and Amazon has filed an amended complaint. Additionally, anti-bot systems designed to prevent fraud are triggering blocks on legitimate AI agents, and web infrastructure provider Cloudflare introduced a one-click block for AI crawlers.
Context
The blocking campaign reveals a fundamental tension between AI companies' vision of ubiquitous autonomous agents and websites' business interests and security concerns. E-commerce platforms face reduced user engagement and ad impressions when agents handle transactions, as Perplexity's case demonstrated. Some companies like Walmart, Gap, Sephora, Best Buy, and Shopify have opted to partner with AI agents rather than block them, suggesting the issue is negotiable but not uniformly resolved. The more websites that block agents, the less capable they become at their core promise—performing any online task on users' behalf—which could materially affect how users perceive AI agents' value. The proliferation of anti-bot defenses may also make routine web use more cumbersome for humans, as sites deploy more CAPTCHA-style verification to distinguish legitimate users from automated visitors.